Left Coast Views
columnist: William Westmiller

Topic: Presidential Campaign 2008
Real Clear Politics or Very Fuzzy Fables?

---

The horserace among presidential candidates is exciting and fun, but the contest is for hearts and minds. The polling numbers are almost pure fantasy.

---

by William Westmiller
(libertarian)
Thursday, January 3, 2008

Admit it: you just love the horserace. One candidate up, another candidate down; every day new numbers; almost as though they had some magical insight into the changing minds of millions of voters. Headlines every other day describe a "surge", "boomlet", "sag", or "constant" standing as the candidates round the far turn and head for the finish line. We crave the next poll numbers, the sleek lines jump and drip, while the candidates explain why that last ad, or expensive haircut, or debate blooper, caused their standing to rise or fall. It’s an engrossing story that is mostly fable.

The most popular polling site is Real Clear Politics, which offers running trend lines on colorful graphs and the latest results from the premier pollsters. There are competitors, like Pollster with fancy scatter-charts, US Election Polls with itemized tables, and Polling Report with source data. But, the all pale in comparison with RCP, which has become the media resource and reference for really hard numbers derived from multiple polls.

Yes, I’m picking on RCP. But, I don’t want to imply that anything on their report is false, incorrect, or intended to mislead. They’ve done a marvelous design job and present accurate information from all the top polling firms. What I want to point out is what you do NOT see, unless you wade into the depths of the original polling data reports. For those who don’t know how public opinion polling works, here’s a lengthy primer. For the remainder of this article, you’ll find documentation there for all of my assertions.

The nice crisp lines on the RCP charts show the median number average from the most recent polls. What it does not show is the margin of sampling error for those polls, which can vary from three to five percent. That’s a plus-or-minus number that applies to each of the polls. The scientific statistical fact is that the actual projected number could be anywhere in that range. All of the numbers above and below are equally true, with a 95% confidence level. Now, if RCP decided to show the entire range for each of the candidates, there would be an entirely different picture, shown below. Much more fuzzy and uncertain.

I’ve modified their graph for the Iowa GOP caucus numbers, as reported yesterday on their site. That election is in progress as I write, so the results will be interesting. Notice that the scientifically correct, fuzzy graph still suggests some trends over long periods. I don’t doubt that they are nearly correct, but there’s no way to verify any numbers until everyone is done voting. The trends in that revised chart just aren’t the "headline grabber" story that the superficial "hard numbers" of the pollsters and RCP present. [Click the image for full size.]

That’s only half the problem. When RCP averages multiple polls, we assume that their number is more a more accurate reflection of the "real picture" of voter sentiment. That isn’t true.

Rather than mitigating the margin of sample error, the new number is actually less accurate. That’s because you have to add up all of the error margins of each of the polls in order to accurately reflect the potential range of "true" results. In this chart, I’ve taken one set of numbers for Mitt Romney, which averages four polls. But, each of those polls has a margin of error and that error stretches over a full ten percent range (or more) of the graph. The possibly correct number is somewhere in that broad range. The earlier chart is crystal clear in comparison with what it would look like with the actual, compounded error range. It would be a smear of color bars that would tell you almost nothing.

We’re still not finished.

The pollsters try their best to get answers from participants. Sometimes they report the percentages of those who report whether or not they are committed, leaning, inclined, or tending toward a particular candidate. Most professional pollsters also report the percentage of undecided voters, but many don’t. All of them know that early polls are almost entirely name identification. Only in the final months of a campaign do favorable and unfavorable impressions begin to influence voters. All of the pollsters know, from academic studies, that forty percent of voters don’t decide until the last few weeks of a campaign. Nevertheless, they "give" each of the candidates a portion of an imaginary 100% of the vote (if it were held that day), based on the responses of participants. That isn’t a true picture of where the race stands.

This chart is derived from the RCP graphic, but "stretched and skewed" to show the residual support bars for the candidates, above the likely undecided portion of all those who will actually vote. This is a true, accurate, and scientific graph of what the polls are actually showing us. It’s not just fuzzy, it has the appearance of being a barely educated guess or a very bad abstract painting. But, that’s all the scientific polls are: a rough approximation of sentiments with a multitude of sampling and other errors that produce a single "hard number" for each candidate that is essentially a fantasy.

I don’t expect the pollsters or RCP to change their practices because I’ve shown you the real picture. They make lots of money because you enjoy the horserace and read the stories that are spun around the latest "hard numbers." But, compare the graph above with the original that you’ll find at the RCP site. Check out the detailed reports from the pollsters, then tell me that the original RCP charts are the real picture of where the race stands on any day over the past year. It may be a "best guess," but I call it a fable.

...

Update: I've added the actual Iowa results to the top graphic. Here are the final Iowa polls, with the actual results at the top.

Related: Making Election Polls and Sausage

Did you like this article?
If you did, Thumb It!
4 thumbs so far

©2008 William Westmiller, all rights reserved. You must have written permission from the author in order to republish this work.
Published: Thursday, January 3, 2008
Last modified: Friday, January 4, 2008

The views expressed in this article are those of William Westmiller only and do not represent the views of Nolan Chart, LLC or its affiliates. William Westmiller is solely responsible for the contents of this article and is not an employee or otherwise affiliated with Nolan Chart, LLC in his/her role as a columnist.

Report violation by William Westmiller of Nolan Chart LLC's terms of use policy.